An Image Semantic Segmentation Method Based on Adversarial Training

Abstract

The invention belongs to the technical field of computer vision and discloses an image semantic segmentation method based on adversarial training, which is used to solve the problem that the existing semantic segmentation method cannot realize the meaningful fusion of image global information and local information and cannot learn high-level information in images. The question of potential. Since the present invention defines the loss function of the entire network based on the adversarial training network, the adversarial network as a general function approximator can not only learn how to combine information of different layers, but also "force" the generation network to learn a single point in the segmented picture , pairing, high-order potential energy and other information, realize the organic fusion of image local features and all features, and obtain more realistic segmentation images; at the same time, the method of layer-by-layer training avoids the complicated process of network initialization parameters, so that the entire network can be used Method of random initialization.

Description

technical field [0001] The invention belongs to the technical field of computer vision and relates to image semantic segmentation and confrontation training, in particular to an image semantic segmentation method based on confrontation training. Background technique [0002] With the development and popularization of artificial intelligence, the important position of image semantic segmentation in the field of computer vision has become increasingly prominent. Many applications require accurate and efficient segmentation techniques, such as autonomous driving, indoor navigation, human-computer interaction, and more. In the past five years, deep learning methods have achieved great success in the field of computer vision, and various network structures have been proposed to solve different problems in this field, such as image classification and positioning. However, the most successful network structure convolutional neural network (Convolutional Neural Network, CNN) in the...

AI Technical Summary

Problems solved by technology

However, the most successful network structure convolutional neural network (Convolutional Neural Network, CNN) in the field of computer vision has a major disadvantage for image semantic segmentation: due to the large number of maximum pooling layers stacked in the network structure, CNN finally obtains The feature is the information of the whole picture, while ignoring the local features of the picture, such as the edge and position of the object in the picture

[0005] (1) The cross-layer connection is too simple for the fusion of different layers of information, and cannot realize the meaningful fusion of global information and local information of the image

Images